cv

This page contains my latest CV.

General Information

Full Name Shivesh Prakash
Email shivesh_prakash@berkeley.edu
LinkedIn linkedin.com/Shivesh777
GitHub github.com/Shivesh777

Education

  • 03/2027
    Master of Financial Engineering
    University of California, Berkeley - Haas School of Business
    • Anticipated Coursework: Derivatives, High-Frequency Finance, Fixed Income, Stochastic Calculus with Asset Pricing.
  • 05/2025
    Honors Bachelor of Science
    University of Toronto
    • Major: Computer Science (AI, CV), Statistics Minor; GPA: 3.88
    • Relevant Coursework: Machine Learning(A+), Graduate Optimization(A), Probabilistic Learning(A), Math of Finance(A+).
    • Awards: International Scholar ($100,000), DSI Award ($7,200), JACP Scholar ($500), UC Merit Award, 3x Dean’s List.

Professional Experience

  • 09/25 – Pres.
    Summer Intern (Treasury Quant and Risk Analytics)
    Ernst and Young, Mumbai, India
    • Engineered a privacy-first RAG query engine for corporate treasury, contributed to securing 4 new clients.
    • Building an API pipeline to automate fund allocation & order execution, cutting 2 hrs of manual effort per trade.
  • 06/25 – 09/25
    Quantitative Research Intern
    Pace Stock Broking Services (HFT desk), Delhi, India
    • Built ML models for alpha generation and execution optimization on Indian equities in an HFT environment.
    • Developed a custom clustering pipeline (PCA + HDBScan) on minute-wise returns to group co-moving stocks.
    • Achieved 3.3% auto-corr on vol-weighted sector returns, supporting portfolio management & alpha strategies.
  • 06/24 – 08/25
    FinHub Research Assistant
    University of Toronto
    • Built an RL environment simulating informed, passive. & market maker interactions for the LT2 case on RIT.
    • Processed 9TB of WSJ/NYT Bloomberg data with RBC NY to extrapolate indices and study credit spreads.
    • Used symbolic diff in Python to optimize household investment models under constraint systems.
    • Modeled volatility surfaces to study the impact of non-standard preferences on options (SPP) pricing.
    • Designed LeftNet-based GNN for transition state prediction (18.6 pm MAE); won Best Poster at DSI SUDS.

Research Publications & Papers

  • VIXtrader
    Supervisor- Prof. Ing-Haw Cheng | To be submitted to JFDS (October 2025)
    • Forecasted 34-day VIX using LSTM on macro + ARMA features; achieved 10% lower MAE and 2x R2 vs baseline.
    • Built & backtested dynamic leverage strategies; achieved Sharpe ratios 2.6x higher than passive benchmarks.
  • MHNpath
    Supervisors- Prof. H.A. Jacobsen, Prof. V.K. Prasad | Under review at Digital Discovery Journal
    • Developed an A∗-inspired tree search framework using Hopfield Networks for retrosynthetic pathway prediction.
    • Achieved 89.1% top-1 and 99.6% top-100 accuracy; predicted viable routes for 60% of novel compounds.
  • SEE-2-SOUND
    Supervisor- Prof. Houman Khosravani | Accepted to ICML’24 W and SIGGRAPH’25 P
    • Introduced the first zero-shot, training-free pipeline for generating 5.1 spatial audio from visual content.
  • Material Sensing
    Supervisors- Professor Alex Mariakakis, Dhruv Verma
    • Developed low-latency sub-second hyperspectral capture & DNG-RGB demosaicing pipeline; trained ML models.

Skills & Certifications

Technical Skills Python, C/C++, CUDA, SQL, R, Matlab, Arduino, TensorFlow, PyTorch, NumPy, Pandas, Scikit Learn.
Certifications CFA Level 1, Akuna Capital Options 201 & 101, ML and DL Specialization by Stanford University.

Achievements

  • World Top 10 at Akuna Capital's 2025 Quant Trading Challenge
  • Top 100 in NK Securities Hackathon
  • JEE All India Rank 389
  • Physics Olympiad State Topper & National Top 1%
  • 2x Regional Math Olympiad Qualifier

Projects

  • NanoTick LOB Replay Engine | C++20, Parquet, PyArrow/Numba
    • Replays Nasdaq BX ITCH 30M messages in 7s (4.3M msg/s) on Mac M1, with p99 = 6ns for all book operations. Loss-free ITCH 5.0 to Parquet pipeline ingests 36M BX messages in 128s (280k msg/s).
  • DygnosTech | Python, Keras, OpenCV, Tensorflow, Web Assembly, Streamlit, Twilio
    • Built a GAN-based system for drug side-effect prediction with WebAssembly-optimized models. Won MetroHacks ($4000).

Extracurricular Activities

  • June'23 - Apr'25
    Founder
    UofT Computer Vision
    • Built a community of 120+ members, secured $4220 in funding.
  • Oct'22 - Sep'24
    Machine Learning Project Lead
    UofT Aerospace Team
    • Led projects resolving keystone distortion and dark noise in hyperspectral imaging.

Other Interests

  • Anything finance: trading, derivatives, market microstructures
  • Sports: Cricket, Chess, Badminton
  • Artists: Rahat Fateh Ali Khan, Avicii